Evidence and retention
Window boundaries are real
Section titled “Window boundaries are real”An incident’s SLA window runs from quorum confirmation to confirmed recovery, the same boundaries you were paged on. An exclusion window derived from a discarded false positive is computed per affected region: from the last good check before the incident to the first good check after it. Nothing is stretched to a flattering minute.
What counts, and what is deliberately left out
Section titled “What counts, and what is deliberately left out”An incident enters the report if it overlaps the reporting period. Two classes stay out:
- Excluded incidents - a downtime you curated away as a false positive. The exclusion is a deliberate decision, and letting it also count in the figures would apply it twice.
- Deleted incidents.
Announced maintenance windows that affect the monitor are carried in the report as their own class, not silently folded into availability. A reader can see what was planned and what was not.
What the timeline shows - and what it does not
Section titled “What the timeline shows - and what it does not”An incident’s timeline in an SLA report contains only events explicitly marked
public. Event visibility defaults to internal, so an event reaches an
outside reader only when somebody released it.
That filter sits in the query that builds the report, not in the endpoint that calls it. The distinction matters more than it looks: today the endpoint is read by members of the organization, who may see everything anyway - so a missing filter would not show up in testing. But a report exists in order to leave the house. It goes to a customer, to an auditor, later possibly as a PDF or through a narrowed API key. It must never carry more than what may go outside, and the safest place to guarantee that is the query itself.
Practically: if you want an internal note to appear in customer-facing evidence, mark it public deliberately. Nothing leaks by default.
The math
Section titled “The math”Availability is floored, never rounded up: an SLA figure should be a lower bound you can defend. The sample SLA report shows the full structure with numbers that add up; hand it to anyone and let them recompute it.
What each history window controls
Section titled “What each history window controls”The plan windows are 7 days (Free), 30 (Pulse), 90 (Sentinel), 365 days (Command), and 730 days (Enterprise). The implementation applies them by object, not as one blanket retention period:
| Data or view | Implemented horizon |
|---|---|
| Public history of resolved monitor incidents | Limited to the plan window. This is a read limit, not automatic deletion of the incident record. |
| Raw external probe checks | Physical pruning uses the plan window plus a 2-day technical buffer. |
| Raw host-agent measurement curves | At most 90 days, including on plans with a longer window. |
| Status-page uptime history | At most 365 days, including on Enterprise. |
| Daily rollups, incident and post-mortem versions, activity, and other records | Object-specific lifecycle. The plan row alone does not define their deletion, and some currently have no automatic expiry. |
The current Terms separately state a 13-month standard for measurement and event data. That wording does not match this object-specific implementation, and legal reconciliation is still open. This page records current product behaviour; it does not replace binding terms. Confirm the applicable retention terms through the security review before procurement or deployment.
Export
Section titled “Export”Perstat currently provides an organization export as CSV plus manifest manually on request; self-service report download is not live. For an exit request and the current deletion process, use the route described on the security page.
What this is, and is not
Section titled “What this is, and is not”Perstat can contribute system timestamps, measurements, maintenance, and visible curation to the wider work behind an assurance report or DORA notification. It is not an assurance report, regulatory filing, or compliance decision; the evidence page for leadership draws that boundary precisely.